OpenAI plans to buy artificial intelligence (AI) computing chips from Google as it looks to increase its computing capacity and strengthen its AI infrastructure, according to Bloomberg, which cited a person familiar with the matter.
The final terms of the possible deal have not yet been decided, the person said. The person spoke on the condition of anonymity because the discussions are private.
OpenAI looks to increase AI computing power
The planned purchase is part of OpenAI’s efforts to secure more computing power to develop and run its AI models. The company is looking at different ways to increase its computing capacity instead of depending on a single source of chips, according to Bloomberg’s report.
OpenAI has also announced plans to start using its own AI chips to support its AI models later this year. This shows that the company is working on multiple ways to meet its growing need for computing power.
OpenAI develops its own AI chips
OpenAI is among several technology companies developing their own AI chips to reduce their dependence on Nvidia, which currently dominates the market for AI computing hardware, according to Bloomberg.
{{/usCountry}}OpenAI is among several technology companies developing their own AI chips to reduce their dependence on Nvidia, which currently dominates the market for AI computing hardware, according to Bloomberg.
{{/usCountry}}Also read: S&P 500 investors face new risks as Spacex-Tesla merger, Anthropic IPO plans emerge
OpenAI’s in-house chips, called Jalapeno, are being developed with the help of Broadcom. The project also has links to Google’s chip development work, as people who previously worked at Google have helped with the effort, Bloomberg reported.
Richard Ho, who leads OpenAI’s internal chip team, was previously involved in developing Google’s semiconductors. Google’s AI chips are known as tensor processing units, or TPUs, Bloomberg reported.
Google TPUs offer an alternative to Nvidia chips
Google began developing TPUs more than a decade ago. The company initially designed these processors to speed up its web search engine and make its operations more efficient, according to Bloomberg.
Over time, Google’s TPUs became useful for handling AI-related tasks. They have emerged as an alternative to Nvidia’s graphics processing units (GPUs), which are widely used to train and run AI models, Bloomberg reported.
Oracle to offer Google AI chips through cloud services
Google’s chips are also becoming available through other companies’ cloud services. Oracle announced on Tuesday that customers would be able to access Google’s TPUs through its cloud-computing infrastructure.
Announcing the Oracle deal, Google executive Amin Vahdat said that future AI systems would need different types of computing power. “The next generation of AI will require a diverse range of compute,” Vahdat said in a statement.
The reported OpenAI deal highlights the growing competition among technology companies to secure the chips and computing infrastructure needed to develop more advanced AI systems. It also reflects the wider industry effort to find alternatives to Nvidia’s hardware.